[cfp-interest 3384] JTC1/SC22/WG14/CFP 2025/02/12 Meeting minutes

Thu Feb 13 12:04:26 PST 2025

2025/02/12: 8:00 AM PST
[Please submit proposed changes to these minutes to Jerome or to the group.
Revision changes appear at the bottom.]

*Attendees*

    Jim Thomas, Rajan Bhakta, Jerome Coonen, Fred Tydeman, Joshua Cranmer,
Damian McGuckin, David Hough, Mike Fairhurst, Tue Ly

*New agenda items*

   *https://wiki.edg.com/pub/CFP/WebHome/CFP%20meeting%20agenda-20250212-update.pdf
<https://wiki.edg.com/pub/CFP/WebHome/CFP%20meeting%20agenda-20250212-update.pdf>*

    SIGFPE

*Previous meeting notes*

https://wiki.edg.com/pub/CFP/WebHome/CFP%20Meeting%20notes%202025_01_08%20--%20Rev.%202.pdf

*Study group logistics*

    Next meeting: 12 March 2025, 8:00 AM PST

    ISO Zoom teleconference

     Please notify the group if this time slot does not work.

     Jim: David has been fighting DDOS attacks on the servers he provides
to support CFP.

*C documents*

    There is a new C2Y draft 3435 Jan 2025.
https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3435.pdf

    C23 has been published ISO/IEC 9899, available for purchase.
https://www.iso.org/standard/82075.html

*IEEE 754 liaison*

    Jerome & Damian: Approved improvements to 5 definitions. Rejected a
proposal re. signs of NaNs and their ordering. Discussed removal of
"outdated" items -- extended formats in the mix. Jim Demmel is leading an
effort to look at complex arithmetic.

Tue: Google is working on complex functions corresponding to the math
functions in math.h.

Jim, et al.: Although the imaginary type is in C23, Tue reports that the
discussion to remove the imaginary type from the C2Y working draft has
caused them to back off on their implementation. So the circle completes,
because lack of implementation is the reason for WG14 to remove imaginary
from the standard.

*C++ liaison*

    Joshua: Meeting in progress in Austria. FP on agenda maybe tomorrow...

*WG14*

    Next meeting is Feb 24-28.

    Several items in the discussion that  follows refer to topics Rajan
and/or Fred will bring to the meeting – *due to Rajan by Wed. 19 Feb.*

*C23 integration*

    None.

*Carry-over action items from last meeting *

    David: Will draft reply for Rajan to take to WG14 re midpoint and
interpolation. The idea is that, generally, it is not a good idea to put
extensive library support into languages (e.g. LAPACK). Midpoint and
interpolation are complicated, especially in decimal, and are beyond what
anyone would ever put in hardware.

    Damian: Split doc into 3 documents: complex.h intro, cproj, and
editorial suggestions
        Damian: cproj -- have worked with Jerome on proposal to be put on
wiki; editorial suggestions -- working with Jerome; complex.h intro --
coming.
        Jim: Will keep as a carry-over item.

*Action items from last meeting *

    Damian: Prepare cproj proposal for WG14. Done.

    Jim: Submit [Cfp-interest 3337] related to complex literals to WG14.
Done

    N3452 2025/01/12 Thomas, Proposal for C2Y - Complex literals warning.
Done.

    All: Continue study of the imaginary I macro and i suffix issues,
noting that the I macro may be removed from C. Ongoing.

*TS-4 and TS-5 revisions*

    Jim: Have been published, completing the task of IEEE 754 bindings.

    Rajan: There have been comments from WG14. Would like to haver feedback
by 22 Feb.

*Discussion of issues*

   Comment on CFP proposal
        [cfp-interest 3369] Fwd: [SC22WG14.28504] 2024/12/01 Thomas,
Proposal for C2Y - Improved treatment of error conditions for functions
that round result to narrow type Jim Thomas
        Jim, Jerome: Feedback about unclear wording inspired a re-do of the
document with more explicit language. That language raises the spectre of
unsigned infinities in fsub.
        Group: Consensus is to step lightly on unsigned infinities, saying
as little as possible to cover the case of INF-INF. Jim proposes that we
tweak all 5 items to achieve more parallel structure.
        Rajan: Want something for the WG14 meeting. (New action item
entered.)

    cproj cleanup
        Covered in carry-over action items.

    Errors in draft 3435
        [Cfp-interest 3360
<https://mailman.oakapple.net/pipermail/cfp-interest/2025-January/003375.html>]
Errors in N3435 changes Jim Thomas
        [Cfp-interest 3361
<https://mailman.oakapple.net/pipermail/cfp-interest/2025-January/003376.html>]
Re: [SC22WG14.28105] Errors in N3435 changes Joseph Myers
        Jim: The changes suggested in the 33360 mail were not made
correctly. Mail 3361 requests changes to fix the fiexes. Joseph fixed the
typeface problem.
        Rajan: Worked with WG144 to get the other issue, the parenthetical
comment removal, complete.
        Jim: Finally done.

    CFP proposal updates
        [Cfp-interest 3362
<https://mailman.oakapple.net/pipermail/cfp-interest/2025-January/003377.html>]
CFP proposal updates Jim Thomas
        Jim: Have added comment for future directions regarding the
obsolescent feature. Already submitted.
        Fred: Would it help to note that the obsolescent range error has
been replaced by a domain error?
        Jim: (After further discussion and wordsmithing.) We will leave as
is and consider Fred's suggestion.

        Jim: There is an inconsistency in the last two bullets, namely, the
extra “, the difference". The fix is to remove the extra text.
        Fred: Looks good.

    Reply to WG14 re midpoint and interpolation
        Covered above.

    Imaginary I macro and i suffix issues
        [cfp-interest 3366
<http://mailman.oakapple.net/pipermail/cfp-interest/2025-February/003381.html>]
complex suffix specification Jim Thomas
        Jim: Suggest paragraph break to split suffix from next para. Also,
fix wording to avoid specifying a type the implementation does not provide.
        Group: An existential crisis seems to arise because, on a system
supporting decimal floating types (DF), there "can't" be complex decimal
floating types because the standard doesn't define any. The suffix DFI is
suggestive, but actually doesn’t designate a type at all since C doesn’t
specify decimal complex types.
        Rajan: Suggest paper to expose this to everyone. In a system
without decimal, DFI suffix specifies a type that is not supported, so it
must lead to a diagnostic. The fact that it COULDN'T be support is
irrelevant. Whether there is a constraint violation if decimal is supported
might depend on interpretation.
        Jim: (Following some wordsmithing.) Let's leave this for further
discussion.

        [cfp-interest 3368
<https://mailman.oakapple.net/pipermail/cfp-interest/2025-February/003383.html>]
removing imaginary I Jim Thomas
        Jim: Suggest 1.0fI for implementations that want the I macro.
        Damian & Jim: As a note vs. footnote.
        Jim: The I macro is used in cproj, but we are about to replace that
with a much improved version that omits the I macro. Let's leave cproj as
is, pending that proposal.
        Rajan: Will mention all these issus in the WG`14 meeting.

    SIGFPE
         https://wiki.edg.com/pub/CFP/WebHome/n3378c.htm
         Fred: Joseph Meyers prefers to avoid implementation-defined in
options 3 and 4. He prefers option 1, which is my choice. There is no
indication of breaking anything.
        Tue: Functions receiving signaling NaNs might raise SIGFPE.
        Fred: Standard C does not allow SIGFPE -- you should be raising
Invalid Operation in 60559.
        Rajan: So we are settling on option 1.
        Fred & Tue: Prefer option 1.
        Jim: Note this is at the edge of CFP. It's being presented as a C
issue, not a CFP issue.

    CFP accomplishment and future
        [cfp-interest 3374
<http://mailman.oakapple.net/pipermail/cfp-interest/2025-February/003389.html>]
CFP accomplishment and future Jim Thomas
        [cfp-interest 3377
<http://mailman.oakapple.net/pipermail/cfp-interest/2025-February/003392.html>]
Re: CFP accomplishment and future Damian McGuckin
        [cfp-interest 3378
<http://mailman.oakapple.net/pipermail/cfp-interest/2025-February/003393.html>]
Re: CFP accomplishment and future Damian McGuckin
        [cfp-interest 3379
<http://mailman.oakapple.net/pipermail/cfp-interest/2025-February/003394.html>]
Re: CFP accomplishment and future Mike Cowlishaw

        Jim: C23, TS5, and TS6 cover the goal of CFP. Started in 2009. C23
was the first major revision, so finally chance to get the bindings.

        Rajan: Great chance to report to CFP. For future, do we want to
address fixed point.

        Tue: Google has started fixed point, matching the fixed point TS.

        Rajan: IBM also has a fixed point implementation.

        Joshua: Fixed point types and other fixed point TS issues are tied
to embedded implementations.

        Group: Fixed point falls outside the floating domain. This merits a
new working group with a new charter.

        Jim: Mike Cowlishaw has suggested considering moving some C
floating-point features to annexes or TSes or removing them altogether.
There is sentiment for similar changes in 754.

*Other issues:*
    Correctly rounded math functions

        [cfp-interest 3372
<https://mailman.oakapple.net/pipermail/cfp-interest/2025-February/003387.html>]
Accuracy of Mathematical Functions Paul Zimmermann

*Adjournment*

    10:08 AM PST

-----------------------------------------------

*Action items to be carried over:*

    Damian: complex.h intro, cproj, editorial suggestions.

*New action items:*

    David: Draft statement on minpoint and interpolation.

    Jim: Review the comments on TS 5.

    Jerome: Rework the proposal for functions rounding to a narrow type.
Finish by 19 Feb for Rajan to take to WG14 meeting.

    Jim: Write a paper proposing editorial changes for imaginary suffix,
emphasizing editorial changes vs. technical changes.

    Fred: Prepare SIGFPE proposal based on WG14 feedback and CFP discussion.

*Discussion issues to be carried over:*

    None.

Respectfully submitted.

-Jerome Coonen
 650.996.4738
 jcoonen at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20250213/ce44809e/attachment-0001.htm>